AR antagonists develop drug resistance through TOMM20 autophagic degradation-promoted transformation to neuroendocrine prostate cancer

Human PCa cells LNCaP, VCaP, PC-3, DU145, NEPC cells NCI-H660,HEK293T cells and non-malignant prostate epithelial cells RWPE-1 were purchased from ATCC (Manassas, VA, USA). C4–2 and CWR22Rv1 were obtained from Dr. Chinghai Kao at the Indiana University School of Medicine. PCa cells were maintained in RPMI-1640 medium (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% FBS (Gibco, ThermoFisher Scientific, Friendship, ME, USA) and penicillin/streptomycin antibiotics. HEK293T cells were cultured in DMEM (Thermo Fisher Scientific, Waltham, MA, USA,with/without Sodium Pyruvate) medium and supplemented with 10% FBS (Gibco, ThermoFisher Scientific, Friendship, ME, USA) and penicillin/streptomycin antibiotics.RWPE-1 cells were maintained in defined Keratinocyte-SFM (1 ×) liquid (Invitrogen, Carlsbad, CA, USA). NCI-H660 cells were maintained according to the manufacturers’protocols.All cells were cultured with 5% CO2 in a 37 °C incubator.

The lentivirus system was applied to introduce the shRNAs or cDNAs into PCa cells. HEK-293 T cells were co-transfected with the virus package plasmids psAX2, pMD2G, and the gene expression plasmids. After 48 h transfection, the supernatants consisting of viral particles were collected for immediate use and/or frozen at -80℃ for later use.

pcDNA3.1-His/Myc-TOMM20 and pCMV-AR-Flag plasmids were constructed as follows. The cDNA fragments of TOMM20 and AR were generated individually using PCR and then subcloned into the described vectors. The constructed DNA fragments were confirmed by sequencing. Plasmids pcDNA3.1-His/Myc-TOMM20(∆TPR) and pcDNA3.1-His/Myc-TOMM20(∆Glu) were purchased from Genscript ProBio (Nanjing,China).

siRNA transfection was performed using the Mirus transfection kit according to the manufacturer’s instructions (Mirus, Madison, WI, USA). Non-targeting (NT) siRNA was used as the control. The sequences of siRNA (5′-3′) are listed in the Supplementary Table 1.

Total RNA was extracted using TRIzol (Takara Bio Inc.) and 1 µg of total RNA was used for cDNA synthesis using the Super Script III First Strand Synthesis System (Invitrogen) according to the manufacturer’s instructions. Quantitative real-time PCR (qRT-PCR) was conducted using a Bio-Rad CFX96 system with SYBR green to detect the mRNA expression level of a gene of interest. Expression levels were normalized to GAPDH. The primers used for the genes of interest are listed in Supplementary Table 1.

Cycloheximide (CHX) and Bafilomycin A1 were purchased from Cell Signaling Technology (CST). Hydroxychloroquine Sulfate(HCQ), Enzalutamide (MDV3100) and Bicalutamide(Casodex) were purchased from Selleck, China (Shanghai, China). MK2206 was purchased from MedChemExpress (MCE, Shanghai, China). MG132 was purchased from Targetmol,USA (T2154).All were dissolved in dimethylsulfoxide (DMSO) except for HCQ dissolved in water.

AR(rabbit monoclonal, 1:1000, #5153), PSA/KLK3(rabbit monoclonal, 1:1000, #5365), TOMM20(rabbit monoclonal, 1:1000, #42,406), ATG5(rabbit monoclonal, 1:1000, #12994S), LC3I/II(rabbit monoclonal, 1:1000, #12,741), Akt (rabbit monoclonal, 1:1000, #4691), Phospho-Akt (Ser473)(rabbit monoclonal, 1:1000, #4060), Phospho-Akt (Thr308)(rabbit monoclonal, 1:1000, #13,038), Nanog(rabbit monoclonal, 1:1000, #4903), Sox2(rabbit monoclonal, 1:1000, #3579), ALDH1A1(rabbit monoclonal, 1:1000, #36671S) and VDAC(rabbit monoclonal, 1:1000, #4661) antibodies were purchased from Cell Signaling Technology (CST). NCAM1(rabbit polyclonal antibody, 1:1000, A7913), β-Actin(mouse polyclonal, 1:5000, AC004) and α-Tublin(mouse polyclonal, 1:5000, AC012) antibodies were purchased from ABclonal Technology (Upper Heyford, UK). NSE (rabbit monoclonal, 1:1000, ab180943) and Syp (Rabbit monoclonal, 1:1000, ab184176) were purchased from Abcam (Shanghai, China). ALDH1A1(mouse monoclonal antibody, 1:500,sc-166362), TOMM70(mouse monoclonal antibody, 1:500, sc-390545),AR(rabbit polyclonal, 1:1000, Sc-815X) antibodies were purchased from Santa cruz. E-cadherin (rabbit polyclonal, 1:1000, GTX100443),BRN2(rabbit polyclonal, 1:1000, GTX114650) and N-cadherin (rabbit polyclonal, 1:1000, GTX127345) were purchased from Genetex.

LNCaP and C4-2 cells were maintained in 24-well plates (SORFA, Huzhou, China) with coverslips for 48 h. Cells were fixed with 4% paraformaldehyde for 10 min and permeabilized with 0.1% Triton X-100 for addtional 10 min. Cells were then incubated with AR and TOMM20 antibodies at 4 °C overnight. Cells were washed with cold PBS for three times, and then incubated with the second antibody. The nuclei were stained with DAPI. Images were captured and analyzed with a confocal laser scanning microscope.

Cells were lysed in lysis buffer and proteins (10–30 µg) were separated on 8–10% SDS/PAGE gel and then transferred onto PVDF membranes (Millipore, Billerica, MA). After blocking membranes, they were incubated with primary antibodies, HRP-conjugated secondary antibodies, and visualized using the ECL system (Thermo Fisher Scientific, Rochester, NY).

Extracts for immunoprecipitation were prepared using RIPA buffer supplemented with protease inhibitor cocktails. After centrifugation, 100 µl of the supernatants were prepared as input. The extracts were incubated with the indicated antibodies at 4℃ overnight on a rotator, followed by incubation with protein A/G-magnetic beads (MCE) at 4℃ for 2 h on a rotator. After incubation, beads were washed three times in immunoprecipitation buffer and boiled in 1 × loading buffer. Protein samples were analyzed by SDS-PAGE.

pGEX-4 T-1-AR(FL), pGEX-4 T-1AR(∆NTD), pGEX-4 T-1-AR (DBD) and pGEX-4 T-1-AR(∆LBD) were constructed using the GST gene fusion system according to the manufacturer’s instructions. To produce glutathione S-transferase(GST), GST-AR(FL), GST-AR(∆NTD), GST-AR(∆DBD) and GST-AR (∆LBD)fusion proteins, pGEX-4 T-1, pGEX-4 T-1-AR(FL), pGEX-4 T-1AR(∆NTD), pGEX-4 T-1-AR(∆DBD) or pGEX-4 T-1-AR(∆LBD) was individually transfected into BL21 competent cells. Protein expression was induced by ispropylb-D-1-thiogalactopyranoside (IPTG). GST-AR(FL), GST-AR(∆NTD), GST-AR(∆DBD) and GST-AR(∆LBD) and GST proteins were purified by binding to glutathione-sepharose resins (Sangon Biotech, Shanghai, China). Resins were incubated with cell lysates from LNCaP cells at 4 °C for 12 h on a rotator, and then were washed with the washing buffer four times. Resin-bound complexes were eluted by boiling, and subjected to western blotting.

The cells were plated in 6-well plates (500 cells/well) at 37℃ overnight and then treated with Abiraterone or Casodex. Cell culture media were replaced with fresh growth medium containing drugs every 3 days. The cells were fixed and stained with 0.1% crystal violet solution 7 days after drug treatment. Clones with > 50 cells were counted under the microscope and the survival fractions were calculated as the average number of colonies ± SD of three independent experiments.

Cell viability was assessed by MTT assay. Cells were plated in 96-well plates with a cell density of 5 × 10³ cells per well, and incubated at 37℃ for 0, 24,48, 72, and 96 h. The culture medium was replaced with medium containing MTT, and the cells were incubated at 37℃ for 2 h. Cell viability was evaluated by measuring the absorbance of each well at 570 nm with a VersaMax™ microplate reader. Cell invasion assays were performed using a Transwell Boyden chamber (Corning) precoated with matrigel.

All procedures were approved by the Central South University Institutional Animal Use and Care Committee. All 6 to 8-week old male nude mice were subcutaneously injected into the flank with 1 × 10⁶ tumor cells (LNCaP sh-CTR, LNCaP sh-TOMM20; VCaP sh-CTR, VCaP sh-TOMM20) in 100 µl of a 1:1 solution of medium and Matrigel (Corning Biosciences). Eight weeks later, tumor formation rates were calculated for LNCaP cells. When the mean tumor volume of VCaP cells reached 100–150 mm³, mice were divided into four groups and received the following treatments for 3 continuous weeks: vehicle, Casodex (10 mg/kg/d, oral gavage). Tumor size and body weight were measured twice a week, and tumor xenograft volume (V) was calculated using the following formula: V = ab²/2 (a: the long diameter and b: the short diameter). The tumor xenografts were isolated at the endpoint of the experiment, and tumor size and weight was compared by statistical analysis.

To validate the impact of TOMM20 depletion on PCa metastasis, we firstly generated a PC3 luciferase cell line, and then was infected with lentiviruses expressing TOMM20-shRNA, or scramble control shRNA. The stable PC3 luciferase cells with TOMM20 knockdown were generated by drug selection. The cells were injected into NOD/SCID mice via tail veins (1.5 × 10⁶ cells/100 µL). The cancer metastasis was monitored by assessing the bioluminescence at a series of time points by a IVIS Imaging System (PerkinElmer). For the detection of bioluminescence, one mouse was intraperitoneally injected 200 µL of d-luciferin (15 mg/ml, Yeasen, Shanghai, China) before imaging. The photometry of the tumor was calculated by Living Image 3.1.0 software (Caliper Life Sciences), and the data were used to evaluate the degree of cancer metastasis. The mice survival rate was analyzed by using Kaplan–Meier survival curve and log-rank test.

Cells (5 × 10²) were seeded in 96-well ultra-low attachment culture plates (Corning Life Sciences) in MammoCult™ Basal Medium (Stemcell) with 4 µg/ml Heparin Solution (Stemcell) and 0.48 µg/ml Hydrocortisone Stock Solution (Stemcell) for 10–14 days. Spheres were counted under an inverted microscope in triplicate wells.

Mitochondria were isolated from PCa cells and RWPE-1 cells using a mitochondria isolation kit (C3601, Beyotime Institute of Biotechnology, Nantong, Jiangsu, China) according to the manufacturer’s instructions. Briefly, cells were homogenized in ice-cold mitochondria isolation buffer with 1 mM PMSF and centrifuged at 600 g for 10 min at 4 °C. Subsequently, the supernatants were transferred to another centrifuge tube and centrifuged at 11,000 g for 10 min at 4 °C. The sediment was mitochondria. The remaining supernatants were centrifuged at 12,000 g for 10 min at 4 °C to obtain the cytoplasmic fraction. Protein samples were analyzed by SDS-PAGE.

Intracellular ROS levels were measured using a Reactive Oxygen Species Assay Kit (Beyotime Biotechnology, China) according to the manufacturer’s instructions. Briefly, stable cells expressing sh-TOMM20 and sh-Control were incubated with DCFH-DA at 37 °C for 20 min, and then were observed with a fluorescence microscopy (Olympus). The fluorescence was measured at 488 nm excitation and 525 nm emission by a fluorescence spectrophotometer (BD Bioscience).

Total RNA was extracted from LNCaP sh-CTR and LNCaP sh-TOMM20 cells using the TRIzol reagent according to the manufacturer’s protocol. RNA purity and quantification were determined using the NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). RNA integrity was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Then the libraries were constructed using TruSeq Stranded mRNA LT Sample Prep Kit (Illumina, San Diego, CA, USA) according to the manufacturer’s instructions. The transcriptome sequencing and analysis were conducted by OE Biotech Co., Ltd. (Shanghai, China). Reads per kilobase million (RPKM) for each gene were estimated. A gene is considered at a statistically significant level if the folds of expression difference were more than 1.5 between any two samples. The p value was calculated by Cufflinks. Bioinformatic analysis was performed using the OECloud tools at https://​cloud.​oebiotech.​cn.

All data were analyzed by GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA; RRID:SCR_002798) or FlowJo software. Results are presented as mean ± SEM or mean ± SD as indicated. The statistical difference between two samples was analyzed by Students t test. One-way ANOVA was used to analyze the statistical difference of multiple groups. *P < 0.05 and **P < 0.01. P < 0.05 was considered as statistically significant.

Aberrant overexpression of TOMM20 has been reported in several cancer types [27, 31‐36]. By analyzing the GEO database(GSE21034 and GSE80609), we compared the mRNA levels of TOMM20 gene between primary prostate cancer (PCa) and the adjacent benign prostate(BP) or hyperplasia(BPH). The mRNA levels of TOMM20 gene are higher in PCa than BP (Fig. 1A and Fig.S 1A). Since AR is a key molecule for PCa development, we further tested the expression correlation of AR and TOMM20 genes in PCa (the data were from http://​vip.​sangerbox.​com/​home.​html). TOMM20 is positively correlated with AR at the mRNA levels in 548 PCa cases from TCGA database (Person r = 0.5069, Fig. 1B), and at the protein levels in non-malignant prostate epthelial cell line RWPE-1 and PCa cell lines as shown in Fig. 1C. By immunohistochemical (IHC) analysis, we further validated the correlation of AR and TOMM20 protein levels in 6 BPH and 39 PCa cases. AR protein co-localized with TOMM20 in the cytoplasm of PCa cells, and highly and positively correlated. With the elevation of Gleason grade, the protein levels of AR and TOMM20 elevated simultaneously (Fig. 1D, E and Fig.S 1B). These data suggested that in all stages of PCa, the expression level of TOMM20 is closely and positively correlated with AR.

As we knew, 10–17% PCa transformed to NEPC when PCa patients developed resistance to the second-generation AR antagonists. AR signaling reduced or lost and the neuronal-associated genes (such as BRN2, SYP, ENO2) elevated in NEPC tissues [25, 37]. We compared the mRNA levels of TOMM20 gene in PCa(Adeno) to NEPC by analyzing datasets from GEO and cBioPortal. TOMM20 significantly decreased in NEPC (Fig. 1F). The advanced PCa patients received androgen-deprivation therapy(ADT) together with radical radiotherapy. The ten-year biochemical failure-free survival(BRFS) of PCa patients with high TOMM20 level is much greater than those with low level (Fig. 1G). We then analyzed another GEO dataset (GSE8702) in which LNCaP cells were cultured in androgen-free medium for 12 months, which was regarded as a typical approach to promote the transdifferentiation of NEPC [38, 39]. After prolonged androgen deprivation, TOMM20 and the androgen/AR-regulatory genes robustly decreased, while neuronal-associated genes significantly elevated (Fig. 1H). To further validate the results, we cultured LNCaP cells in androgen-free medium for 25 days, and observed that more cells gained neuron-like morphology with more branches. The levels of TOMM20 and AR gradually decreased with the extension of androgen deprivation, while BRN2 and ENO2 were further up-regulated (Fig. 1I). We then compared TOMM20 protein in PCa cell lines C4-2, PC3 and DU145 to the NEPC cell line NCI-H660. The data showed that the protein level of TOMM20 was much lower in NCI-H660 cells than other PCa cells (Fig. 1J). These data indicated that the loss of TOMM20 might play a key role in the transdifferentiation of PCa to NEPC.

It has been reported that AR is located in the mitochondria of PCa cells, and its expression is closely correlated with the malignancy [27]. To validate the cellular localization of AR, we first assessed AR protein levels inside (Mito) or outside (cyt) of the mitochondria in RWPE-1 and PCa cell lines LNCaP and C4-2. The results showed that PCa cells express more AR protein than RWPE-1, and AR is located in mitochondria of PCa cells in addition to cytoplasm (Fig. 2A). In LNCaP and C4-2 cells, AR protein is located in the mitochondria along with high levels of TOMM20 protein (Fig. 2B). Then, we stained AR and TOMM20 with fluorescent antibodies. AR protein clearly co-localized with TOMM20 (Fig. 2C).

We further studied whether AR protein physically interacts with TOMM20. The protein interaction of AR and TOMM20 was verified by co-immunoprecipitation (co-IP) of exogenously expressed proteins in HEK-293 T cells (Fig. 2D). TOMM20 consists of five conserved sub-domains: N-terminal membrane-anchor domain, linker domain rich in charged amino acids, tetratricopeptide repeat motif (TPR), glutamine-rich domain and C-terminal domain. In addition to the N-terminal membrane-anchor domain, the other four domains stay in the cytoplasm (Fig. 2E). We co-transfected the plasmids encoding AR or TOMM20 excluding the TPR/glutamine-rich domains to HEK-293 T cells, and assessed the interaction of exogenously expressed proteins by co-IP. The results identfied that AR interacts with the TPR domain of TOMM20 (Fig. 2F). Meanwhle, the AR proteins with individual domains (Fig. 2G) were mixed with cell lysates of LNCaP cells overexpressing TOMM20, and the protein domains of AR interacting with TOMM20 were identified by GST-pull down assay. The results showed that the LBD and Hinge domain of AR protein physically interact with TOMM20 (Fig. 2H).

The interaction of endogenously expressed AR and TOMM20 proteins was validated in both LNCaP and C4-2 cells in Fig. 2I (line1,3 and 5). Interestingly, AR antagonists such as MDV3100, significantly decreased the protein interaction of AR and TOMM20 (Fig. 2I,line 2,4 and 6). Beyond our expectations, MDV3100, in addition to inducing AR protein degradation(Fig.S 2,A and B), remarkably induced TOMM20 protein degradation(Fig. 2I), but had no impact on TOMM70, another member of translocase of outer mitochondrial membrane family (Fig. 2J). These results demonstrated that MDV3100-induced AR degradation specifically destroyed the protein stability of TOMM20.

We treated LNCaP or C4-2 cells with MDV3100 for 24 h, and found that MDV3100 significantly induced the protein degradation of TOMM20 in addition to AR, but had no impact on the mRNA levels of TOMM20 and AR genes (Fig. 3A,B and Fig.S 3A). However, MDV3100 had no impact on TOMM20 protein stability in AR-negative PC-3 and DU145 cells (Fig.S 3B). The results indicated that AR antagonists induced TOMM20 protein degradation by destroying AR protein stability. Since AR protein interacts with TOMM20, and MDV3100 significantly reduced their interaction, we hypothesized that AR sustained TOMM20 protein stability. AR depletion with siRNA significantly decreased the protein level of TOMM20, but had no effect on the mRNA levels of TOMM20 gene (Fig. 3C and D). We further tested the impact of AR antagonists on the half-life of TOMM20 protein in LNCaP cells. With the inhibition of new synthesized protein by CHX, MDV3100 significantly decreased the half-life of TOMM20 protein, which indicated that AR antagonists accelerated the protein degradation of TOMM20 (Fig. 3E). As we knew, the mechanism of intracellular protein degradation includes the ubiquitin–proteasome system (UPS) and autophagy-lysosomal pathway. We further investigated the regulatory mechanisms by which AR antagonists induced the protein degradation of TOMM20. First, we treated LNCaP or C4-2 cells with the USP inhibitor MG132 or autophagy inhibitors Bafilomycin A1 or Chloroquine (CQ), and identified the recycle mechanism of TOMM20 protein. The results showed that TOMM20 protein degradation mainly through UPS rather than autophagy under physiological conditions (Fig. 3F). Interestingly, the MDV3100-induced TOMM20 protein degradation might be reversed by autophagy inhibitor CQ, but not by UPS inhibitor MG132 (Fig. 3G,H and Fig.S 3C). These results were verified when AR was depleted by siRNA (Fig. 3I). To further verify that AR sustains the protein stability of TOMM20 by inhibiting the autophagy-mediated degradation pathway, we inhibited the autophagy pathway by depleting ATG5 in LNCaP and C4-2 cells, and then treated cells with MDV3100. The results demonstrated that autophagy inhibition significantly prevented MDV3100-induced TOMM20 protein degradation (Fig. 3J). In the effort to further validate the roles of AR in MDV3100-induced autophagy, we assessed the degree of autophagy (ratio of LC3-II/LC3-I) when AR was artificially overexpressed. The rescue of AR expression remarkably decreased the degree of MDV3100-induced autophagy (Fig. 3K). The results clarified that AR antagonists promoted TOMM20 protein degradation through autophagy-lysosomal pathway.

Since AR antagonists decreased TOMM20 protein levels but have no effect on the mRNA level, we want to know the roles of TOMM20 in the development of drug resistance to AR antagonists. We first analyzed the mRNA levels of TOMM20 gene in three pairs of PCa cells sensitive or resistant to AR antagonists from GEO datasets (GSE847). The mRNA levels of TOMM20 gene were lower in drug resistant PCa cells (Fig.S 4A). To explore the exact mechanism, we searched for the most affected pathways by RNA-seq analysis when TOMM20 was depleted with shRNAs in LNCaP cells. The expression levels of ROS-associated genes significantly elevated (Fig. 4A),, suggesting that TOMM20 depletion might elevate ROS levels. Flow cytometry and cellular fluorescence showed that the depletion of TOMM20 significantly elevated the intracellular ROS level in LNCaP and C4-2 cells (Fig. 4C and Fig.S 4B). It is well known that the elevation of ROS levels activates PI3K/AKT, a cell survival signaling cascade [40]. The KEGG pathway analysis showed the depletion of TOMM20 elevated the genes in the PI3K/AKT signaling pathway (Fig. 4B). As we expected, the depletion of TOMM20 remarkably elevated the levels of phosphorylated AKT(p-AKT) in VCaP cells. The antioxidant N-acetyl-L-cysteine (NAC), by decreasing the TOMM20-elevated ROS level, reversely decreased the levels of phosphorylated AKT(Fig. 4D). We further validated that AR antagonist-promoted TOMM20 protein degradation exerted the same effects as TOMM20 depletion. As shown in Fig. 4E and Supplemental Fig. 4C, MDV3100 significantly elevated the intracellular ROS level when it induced TOMM20 protein degradation. The rescue of TOMM20 reversed MDV3100-elevated ROS levels (Fig. 4F and Fig.S 4C), and reversed MDV3100-elevated levels of phosphorylated AKT (Fig. 4G). It has been reported that the activation of PI3K/AKT signaling pathways probably results in the development of drug resistance to AR antagonists [41]. We established two AR antagonist-resistant PCa cell lines, VCaP-Casodex-R (VCaP-C-R) and VCaP-MDV3100-R (VCaP-M-R) [42]. In comparison to the parental VCaP cells, the protein levels of TOMM20 significantly decreased, while the intracellular ROS level, together with the level of phosphorylated AKT markedly elevated in the drug resistant cells (Fig. 4H,I and Fig.S 4D). These results suggested that AR antagonist-induced TOMM20 protein degradation might induce drug resistance by elevating the intracellular ROS level and activating the PI3K/AKT signaling pathways.

We further investigated the effects of TOMM20 depletion on cell proliferation, using clonogenic and MTT assays. Consistently in LNCaP, C4-2, VCaP and CWR22Rv1 cell lines, TOMM20 depletion with shRNAs significantly accelerated cell proliferation (Fig. 5A,B). Using transwell-based assays, we found that TOMM20 depletion significantly promoted the invasion of LNCaP and C4-2 cells (Fig. 5C). We further performed the in vivo animal experiments to validate whether TOMM20 depletion promoted PCa metastasis. We firstly generated a stable PC-3 luciferase cell line with TOMM20-knockdown, and intravenously injected to NOD/SCID mice via tail veins. The non-targeting shRNA was used as the scrambled control. The cancer metastasis was monitored at a series of time points by a IVIS Imaging System. Consistent with the in vitro data, TOMM20 depletion significantly promoted the metastasis of prostate cancer to the abdominal viscera in NOD/SCID mice (Fig. 5E), and the survival rate of mice was remarkably reduced (Fig. 5D). We tested the impact of TOMM20 depletion on the sensitivity of PCa cells to AR antagonists. As shown in Fig. 5F and Supplemental Fig. 5A and 5B, TOMM20 depletion with shRNAs significantly enhanced drug resistance to Casodex in VCaP, C4-2 and LNCaP cells. We further examined the effects of TOMM20 depletion on the anti-tumor effect of Casodex in PCa xenografts. We first established a pair of VCaP cells stably expressing shTOMM20 or shCTR, and induced tumor xenografts in immune-deficient nude mice. The mice were treated with Casodex via o.p. for 21 days. The tumor growth curve and comparison of tumor sizes at the endpoint of the experiment demonstrated that TOMM20 depletion promoted tumor growth and reduced the anti-tumor effect of Casodex (Fig. 5G). Conversely, we tested whether the rescue of TOMM20 might improve the sensitivity of drug resistant PCa cells to AR antagonists Casodex and MDV3100. We elevated the levels of TOMM20 in the drug resistant VCaP cells by transient transfection, and tested the drug sensitivity to Casodex or MDV3100 by clonogenic formation assays. The results showed that the rescue of TOMM20 expression promoted drug sensitivity of the resistant PCa cells to Casodex or MDV3100 (Fig. 5H). The results were verified by CCK8 cell proliferation assay (Fig. 5I). The results demonstrated that TOMM20 depletion promoted cell proliferation and invasion, and further promoted the resistance to AR antagonists.

Bioinformatics analysis of microarray gene expression revealed that the apoptosis-associated genes decreased in TOMM20 depleted cells (Fig. 6A). By performing an Anoikis assay, we found that TOMM20 stable depletion decreased cell apoptosis levels compared to the control cells (Fig. 6B). It has been reported that PCa stem cells (PCSC) play a critical role in the development of therapeutic resistance and subsequent cancer progression [43]. Since TOMM20 depletion remarkably promoted the resistance of PCa cells to AR antagonists, and decreased apoptosis when cultured in ultra-low attachment plates, we tested whether TOMM20 depletion promoted PCa cells to acquire the characteristics of PCSC. CSC markers ALDH1A1 elevated in LNCaP cells when cultured in androgen-free medium for 25 days (Fig. 6C). Gene expression analysis demonstrated the upregulation of CSC-associated genes concomitant with downregulation of TOMM20 in the ENZ-resistant C4-2B cells (Fig. 6D). The depletion of TOMM20 significantly elevated the mRNA or protein levels of such CSC markers as ALDH1A1, SOX2, Nanog and OCT-4 in LNCaP and C4-2 cells (Fig. 6E,F and Fig.S 6A). Compared to parental cells, TOMM20 level was lower but ALDH1A1 level was higher in the drug resistant VCaP cells (Fig. 6G). Since TOMM20 depletion elevated intracellular ROS levels and activated PI3K/AKT signaling pathway (Fig. 4D), we hypothesized that the acquisition of CSC characteristics resulted from AKT phosphorylation. Indeed, AKT inhibitor MK2206 remarkably decreased the level of ALDH1A1 (Fig. 6H). These data validated that TOMM20 degradation activated the PI3K/AKT pathways and promoted the acquisition of PCSC characteristics. We tested the effects of TOMM20 on the CSC characteristics using sphere formation assays. As shown in Fig. 6I and Supplemental Fig. 6B-C, TOMM20 depletion significantly increased the number and size of spheroids, while TOMM20 overexpression conversely decreased the number and size of spheroids in PCa cells (Fig. 6J). We tested the tumor formation capability of TOMM20-depleted LNCaP cells in immune-deficient nude mice. The ratio of LNCaP xenograft formation is ordinarily very low due to the specific characteristics of LNCaP cells. However, beyond our expectations, the ratio of xenograft formation significantly elevated from 1/12 to 5/12 for TOMM20-depleted LNCaP cells, and tumor sizes were remarkably increased (Fig. 6K). These data proved that TOMM20 depletion promoted the reprogramming of PCa cells to cancer stem-like cells.

Recent studies indicated that the transdifferentiation of adenocarcinoma cells into treatment-emergent NEPC (t-NEPC) is a critical mechanism of drug resistance after treatment with the second generation AR antagonists [17], and NEPC cells share the common characteristics of PCSCs [44]. We observed the characters of t-NEPC in two pairs of AR antagonist-resistant VCaP cells. In addition to the decrease of TOMM20 (Fig. 4I), the protein levels of typical NEPC markers ENO2 and NCAM1 remarkably elevated in Casodex or ENZ-resistant VCaP cells compared to parental cells (Fig. 7A). The results were validated by a RNA-seq dataset in a pair of parental and ENZ-resistant C4-2B cells(GSE159548). As shown in Fig. 7B, the mRNA levels of TOMM20, KLK2 and KLK3 significantly decreased, while the NEPC markers NCAM1, ENO2 and CHGA significantly elevated in the ENZ-resistant cells. Interestingly, we observed similar alterations of gene expression in a CRPC and rib-metastasis patient who underwent ENZ treatment, and then developed drug resistance 12 weeks after treatment (PSA re-elevated from 17.7 ng/ml to 29.3 ng/ml) [45]. The RNA-seq data significantly decreased gene expression of TOMM20, KLK2 and KLK3, while significantly elevated NEPC markers NCAM1 and SYN when drug resistance developed ( Fig. 7C).

We further assessed the protein level of TOMM20 in PCa tumor specimens of pten⁻/p53⁻/Rb1⁻ or pten⁻ gene knockout mice by IHC analysis. The PCa tumors of pten⁻/p53⁻/Rb1⁻ gene knockout mice developed to NEPC phenotype [46], and TOMM20 expression significantly decreased in NEPC than PCa tumors of pten- gene knockout mice (Fig. 7D). Additionally, in the tumor specimens of advanced PCa patients, we observed low TOMM20 but high NCAM1 expression in the AR negatve PCa cells(Fig. 7E). RNA-seq data revealed that neuronal-associated genes were up-regulated in TOMM20-depleted cells (Fig. 7F and Fig.S 7A). TOMM20 depletion with shRNA significantly elevated the mRNA levels of NCAM1, CHGA, EN02, SYP and SYN, and elevated the protein levels of BRN2, NCAM1 and SYP in LNCaP and C4-2 cells (Fig. 7G and H).

We demonstrated that TOMM20 depletion elevated the intracellular ROS level and then activated the PI3K/AKT signaling pathway (Fig. 4C and D). TOMM20 depletion significantly elevated the levels of phosphorylated AKT and NCAM1 in LNCaP and C4-2 cells (Fig. 7I). However, the elevated levels NCAM1 were reversed by MK2206, an AKT phosphorylation inhibitor(Fig. 7I). These data validated that AKT phosphorylation is required for the NEPC-transdifferentiation of PCa.

We further validated that TOMM20 depletion promoted NEPC transdifferentiation. LNCaP cells stably expressing shTOMM20 were cultured in androgen-free medium for 30 days. Many cells showed the neuron-like morphology with more branches and longer antennae (Fig. 7J). Additionally, the protein expression of EMT, CSC and NEPC markers was obviously elevated, while the level of PSA decreased (Fig. 7K). Reversely, TOMM20 overexpressions in NEPC cell line NCI-H660 decreased the protein levels of ALDH1A1, SYP, NCAM1 and N-Cadherin, and elevated the level of E-Cadherin(Fig. 7L). The data validated that TOMM20 depletion promoted the NEPC transdifferentiaton of PCa.